Dynamic motion planning algorithm in human-robot collision avoidance

Abstract

The collision-free robotic motion planning algorithm in a dynamic environment is an effective method for the prevention of collision in a human-robot interaction scenario. In this paper, an improvement of the Rapidly-exploring Random Tree (RRT) algorithm is proposed to avoid the potential collision between robot and human by taking the costmap of robot workspace into consideration. The depth camera and Kalman filter are applied to estimate the positions and velocities of dynamic obstacles. The main contribution of this paper is that the spatiotemporal information of moving obstacles is integrated into the robotic motion planning algorithm with the costmap of workspace. Besides, a new costmap generation method is designed based on the velocities of moving obstacles. Finally, a local replanning method is used to accelerate the robotic motion planning algorithm. Experimental results show that the proposed method can efficiently protect human workers from colliding with the robot and the robot can dynamically generate a collision-free path in the human-robot coexisting environment.